A Natural Language-Based Method and System for Automated Evaluation of Bidding Documents

Abstract

This invention relates to the field of bid document management technology, and in particular to an automatic bid document review method and system based on natural language processing. The method includes: identifying target sections; determining historical performance and financial information; judging whether false performance exists; analyzing the complete sentences containing target keywords; calculating material scores; determining price deviation; and determining the review result. This invention achieves multi-dimensional and comprehensive review of bid documents by intelligently analyzing multiple sections such as qualifications, risks, materials, and pricing. This effectively improves the comprehensiveness and depth of the review. Utilizing natural language processing technology, it can accurately identify and extract key information from bid documents, overcoming the subjectivity and limitations of traditional manual review. Quantitative assessment based on risk level significantly improves the accuracy and relevance of risk identification. By comprehensively weighing material scores and price deviation, it significantly improves review efficiency.

Description

Technical Field

[0001] This invention relates to the field of tender document management technology, and in particular to a method and system for automatic evaluation of tender documents based on natural language. Background Technology

[0002] In the field of modern commercial and engineering project bidding, the review of bid documents is a crucial step in ensuring project quality and fair competition. Traditional bid review processes typically rely on manual review, which presents several inherent technical limitations and challenges. First, manual review heavily depends on the personal experience and subjective judgment of review experts, easily introducing unavoidable biases. Second, as bidding projects become increasingly complex, the content and format of bid documents are becoming more diverse and intricate, making traditional manual review methods inadequate to meet the growing demands for technical and information processing.

[0003] In recent years, the rapid development of artificial intelligence and natural language processing technologies has provided new technical pathways for intelligent review of tender documents. These advanced technologies have made automated and intelligent document analysis and evaluation possible. However, most existing technical solutions remain at the level of simple text matching and data comparison, lacking deep semantic understanding of document content and multi-dimensional comprehensive evaluation. In particular, existing technical solutions are still very limited in identifying implicit risk factors and potential problems in tender documents.

[0004] Patent document CN114595661A discloses a method, apparatus, and medium for reviewing tender documents, comprising: performing noise content filtering on the document content extracted from the tender document to obtain a first text of the tender document; obtaining identification features of target chapters associated with the tender and standard features of text information associated with the tender; identifying the first text based on the identification features to determine a second text of the tender document; extracting text information from the second text based on the standard features to obtain the tender material names and tender material values ​​in the second text; performing data alignment on the obtained tender material names and tender material values ​​to obtain tender data of the tender document, so as to generate tender text based on the obtained tender data; and calculating the similarity between the tender data and standard data to review the generated tender text.

[0005] This reveals the following problem: In the existing technology, it is impossible to judge the authenticity and reasonableness of the tender documents, resulting in a low accuracy rate of the review results. Summary of the Invention

[0006] To address this issue, the present invention provides a method and system for automatic evaluation of tender documents based on natural language, which is used to perform risk detection and price reasonableness judgment on tender documents to overcome the problem of low accuracy of evaluation results caused by the inability to judge the authenticity and reasonableness of tender documents in the prior art.

[0007] To achieve the above objectives, the present invention provides a method and system for automatic evaluation of tender documents based on natural language, comprising:

[0008] Identify the target sections in the tender documents to be processed, including the material information section, the tender qualification section, the risk and responsibility section, and the tender price section;

[0009] Analyze the first text information in the bidding qualification section to obtain historical performance information and historical financial information;

[0010] Determine whether there are any false performance records based on the historical performance information and the historical financial information;

[0011] When it is determined that there is no false performance, the second text information in the risk responsibility section is obtained, and it is determined whether there is a target word in the second text information. When it is determined that there is a target word, it is determined whether to analyze the complete sentence where the target word is located based on the number of times the target word appears and the preset standard number of times.

[0012] The analysis results are obtained after analyzing the complete sentence containing the target word;

[0013] When the analysis result is low risk, obtain the third text information in the material information section;

[0014] Determine the number of material feature words and material parameters in the third text information, and calculate the material score based on the number of material feature words, material parameters, and the corresponding number of standard feature words and standard parameters;

[0015] By comparing the material score with the material score threshold, when the material score is greater than the material score threshold, the bid material price in the investment quotation section is determined based on the historical performance information, the historical financial information and the bid material price to determine the price deviation.

[0016] The review result is determined based on the material score and the price deviation.

[0017] Furthermore, the step of determining whether there is false performance based on the historical performance information and the historical financial information includes:

[0018] Calculate the correlation between the percentage fluctuation of sales performance and the percentage fluctuation of revenue data for each consecutive month in the historical performance, and compare the correlation with a preset correlation threshold to determine whether there is any false performance.

[0019] Furthermore, determining whether the target word exists in the second text information includes:

[0020] The text of the risk and responsibility section is segmented into words to obtain several words to be determined;

[0021] By comparing the word to be determined with the preset responsibility keyword database, if the word to be determined exists in the responsibility keyword database, then it is determined that the target word exists in the second text information.

[0022] Furthermore, the step of determining whether to analyze the complete sentence containing the target word based on the frequency of its occurrence compared to a preset standard frequency is as follows:

[0023] Determine the number of times the target word appears in the second text information;

[0024] Compare the number of occurrences with the preset standard number of occurrences to obtain a comparison result;

[0025] Based on the comparison results, it is determined whether to analyze the complete sentence containing the target word. If the number of occurrences is less than or equal to the preset standard number of occurrences, it is determined that the complete sentence containing the target word will be analyzed.

[0026] Further, determining the frequency of occurrence of the target word in the second text information includes:

[0027] Determine the complete sentence in the second text information based on the target punctuation;

[0028] The number of sentences in which the target word appears in the complete sentence is determined to obtain the occurrence count.

[0029] Furthermore, obtaining the analysis results after analyzing the complete sentence containing the target word includes:

[0030] By using natural language processing technology to perform semantic analysis on the complete sentence containing the target word, the object and manner of responsibility can be determined.

[0031] When the entity undertaking the task is the bidding company, a pre-set risk analysis model is used to conduct a risk analysis on the undertaking method to obtain the analysis results.

[0032] Furthermore, the calculation of the material score based on the number of material feature words, material parameters, and the corresponding number of standard feature words and standard parameters includes:

[0033] The feature score is determined based on the number of material feature words and the number of standard feature words;

[0034] The material score is determined based on the material parameters and the standard parameters.

[0035] The material score is calculated by adding the feature score and the material score together with different weights.

[0036] Furthermore, the determination of price deviation based on the historical performance information, the historical financial information, and the price of the tendered materials includes:

[0037] Calculate the ratio of total sales volume in the historical performance information to total revenue in the historical financial information to obtain the sales price;

[0038] Calculate the ratio of the price of the tendered materials to the selling price to obtain the price deviation.

[0039] Furthermore, determining the review result based on the deviation between the material score and the price includes:

[0040] The ratio of the material score to the price deviation is calculated to obtain the review result.

[0041] This invention also provides a natural language-based automatic bid document review system applicable to any of the above-mentioned automatic review methods, comprising:

[0042] The chapter determination module is used to identify the target chapters in the tender documents to be processed.

[0043] The text analysis module, which is connected to the chapter determination module, is used to analyze the first text information in the bidding qualification chapter to obtain historical performance information and historical financial information.

[0044] A fraud detection module, which is connected to the text analysis module, is used to determine whether there is fraudulent performance based on the historical performance information and the historical financial information.

[0045] The risk assessment module, which is connected to the chapter determination module and the fraud determination module, is used to obtain the second text information in the risk responsibility chapter when it is determined that there is no fraudulent performance, and to determine whether to analyze the complete sentence where the target word is located based on the number of times the target word appears and the preset standard number of times when the target word is present in the second text information, and to obtain the analysis results.

[0046] The material scoring module, which is connected to the chapter determination module and the risk assessment module, is used to determine the number of material feature words and material parameters in the third text information and calculate the material score when the analysis result is low risk.

[0047] The deviation calculation module, which is connected to the chapter determination module and the text analysis module, is used to compare the material score with the material score threshold, and determine the price deviation when the material score is greater than the material score threshold.

[0048] The review module, which is connected to the material scoring module and the deviation calculation module, is used to determine the review result based on the material score and the price deviation.

[0049] Compared with existing technologies, the beneficial effects of this invention are that by intelligently analyzing multiple sections of the tender documents, such as qualifications, risks, materials, and quotations, a multi-dimensional and comprehensive review of the tender documents is achieved, effectively improving the comprehensiveness and depth of the review. By using natural language processing technology, key information in the tender documents, such as historical performance, financial information, risk responsibilities, and material characteristics, can be accurately identified and extracted, overcoming the subjectivity and limitations of traditional manual review. By conducting in-depth analysis of keywords in the risk responsibilities section, potential risk points are accurately identified, and quantitative assessments are made based on the degree of risk, significantly improving the accuracy and pertinence of risk identification. By comprehensively weighing material scores and price deviations, the final review result is formed, significantly improving review efficiency.

[0050] Furthermore, by focusing on the dynamic fluctuation trends of monthly performance and revenue, abnormal fluctuation patterns can be captured, providing a more dynamic and comprehensive assessment basis for determining false performance. By comparing the correlation between sales performance fluctuations and revenue data fluctuations, it is possible to more objectively identify whether bidders have deliberately fabricated performance information, effectively revealing potential performance fraud, and improving the accuracy and reliability of bid document review.

[0051] Furthermore, by establishing a professional responsibility keyword database, it is possible to quickly and accurately capture potential risk signal words in the risk responsibility section, effectively overcoming the subjectivity and arbitrariness in manual review, effectively identifying hidden risk points and uncertainties in the tender documents, and using natural language processing technology to perform refined word segmentation of the text, achieving precise deconstruction of the tender documents. It can quickly and automatically screen a large number of documents for risks, significantly improving work efficiency and reducing labor costs.

[0052] Furthermore, by conducting detailed statistics and threshold comparisons of the frequency of target words, risk-sensitive statements can be identified and located more accurately. The risk level can be automatically judged based on the frequency of target words. By filtering through word frequency thresholds, high-risk statements can be quickly identified, avoiding unnecessary risk misjudgments due to the appearance of individual words. This enables the quantification and hierarchical management of risks, significantly reducing the amount of computation and processing time, and improving the overall efficiency of risk assessment.

[0053] Furthermore, by accurately identifying sentence boundaries and counting the number of sentences in which target words appear in complete sentences, we can ensure that the analysis of text based on complete semantic units can more accurately reflect the distribution and importance of keywords in the text. Compared with the traditional word-by-word and paragraph-by-paragraph analysis method, the rapid location of sentence boundaries through punctuation marks significantly improves the computational efficiency of text parsing and risk identification.

[0054] Furthermore, through natural language processing technology, it is possible to accurately identify the subject and method of liability assumption, intelligently identify the specific object of liability assumption, and distinguish the liability boundaries of different subjects. This provides a more accurate subject positioning for risk assessment, avoiding subjective judgment and fuzzy identification in traditional methods. When the object of liability assumption is the bidding company, a targeted assessment is conducted using a preset risk analysis model, achieving accurate quantification and scientific evaluation of risk. This overcomes the limitations of traditional qualitative analysis. Through detailed analysis of the assumption method, risk assessment can be performed quickly and automatically on a large number of documents, significantly improving work efficiency.

[0055] Furthermore, by simultaneously considering the number of material characteristic terms and material parameters, the overall quality of the material can be reflected more comprehensively. By adopting two dimensions, feature scoring and material scoring, and assigning different weights, the scoring weights can be flexibly adjusted according to the characteristics of different industries and projects, which has good adaptability and customization capabilities. It also achieves accurate quantification of material quality, overcomes the subjectivity and arbitrariness of traditional qualitative assessment, and improves the accuracy and objectivity of the assessment.

[0056] Furthermore, by comprehensively considering historical performance, financial information, and current bid prices, the comprehensiveness and accuracy of price review are significantly improved. The method of calculating the ratio of bid material prices to historical sales prices can sensitively capture price deviations, providing a scientific quantitative indicator for identifying unreasonable quotations, effectively preventing potential commercial fraud and unreasonable pricing risks, and providing effective decision support for bid evaluation.

[0057] Furthermore, by dividing the material score by the price deviation, the bid documents can be automatically reviewed based on material quality and price risks, realizing intelligent calculation of bid review results and significantly improving review efficiency and accuracy. Attached Figure Description

[0058] Figure 1 This is a flowchart of an automatic bid document review method based on natural language according to an embodiment of the present invention;

[0059] Figure 2 A flowchart for determining the existence of a target word in an embodiment of the present invention;

[0060] Figure 3 This is a flowchart illustrating the process of determining the frequency of target words in an embodiment of the present invention.

[0061] Figure 4 This is a system framework diagram of the automatic bid document review system based on natural language according to an embodiment of the present invention. Detailed Implementation

[0062] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0063] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0064] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0065] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0066] Please see Figure 1 ,like Figure 1 The diagram shown is a flowchart of an automatic bid review method based on natural language according to an embodiment of the present invention.

[0067] Specifically, embodiments of the present invention provide a method for automatic evaluation of tender documents based on natural language, including:

[0068] Identify the target sections in the tender documents to be processed, including the material information section, the tender qualification section, the risk and responsibility section, and the tender price section;

[0069] Analyze the first text information in the bidding qualification section to obtain historical performance information and historical financial information;

[0070] Determine whether there are any false performance records based on the historical performance information and the historical financial information;

[0071] When it is determined that there is no false performance, the second text information in the risk responsibility section is obtained, and it is determined whether there is a target word in the second text information. When it is determined that there is a target word, it is determined whether to analyze the complete sentence where the target word is located based on the number of times the target word appears and the preset standard number of times.

[0072] The analysis results are obtained after analyzing the complete sentence containing the target word;

[0073] When the analysis result is low risk, obtain the third text information in the material information section;

[0074] Determine the number of material feature words and material parameters in the third text information, and calculate the material score based on the number of material feature words, material parameters, and the corresponding number of standard feature words and standard parameters;

[0075] By comparing the material score with the material score threshold, when the material score is greater than the material score threshold, the bid material price in the investment quotation section is determined based on the historical performance information, the historical financial information and the bid material price to determine the price deviation.

[0076] The review result is determined based on the material score and the price deviation.

[0077] The target section refers to the section of the tender document containing the content to be reviewed during the review process. The historical performance information refers to the monthly sales performance of the bidding company in the tender document. The historical financial information refers to the monthly revenue of the bidding company in the tender document. The target keywords are terms related to risk management, liability exemption, and disclaimer clauses, such as risk, liability, and disclaimer. In this embodiment, it is risk, liability, and disclaimer. The preset standard frequency is the highest number of times the statements regarding the allocation of liability and risk by the bidding company appear in the tender document, generally set at 1-3 times. In this embodiment, the preset standard frequency is 1 time. The material characteristic terms are beneficial material characteristics of the tendered materials in addition to the essential key characteristics, generally such as corrosion resistance, high strength, and environmental protection. In this embodiment, it is corrosion resistance. Corrosion resistance, high strength, and environmental friendliness are the key characteristics of the tendered materials. These parameters are typically hardness and strength, and in this embodiment, hardness. The number of standard terms represents the minimum number of beneficial material characteristics, generally set between 1 and 4, and in this embodiment, it is 3. The standard parameter represents the minimum value of the key characteristic, generally set between 50 and 100 HB, and in this embodiment, it is 80 HB. The material scoring threshold is the minimum material score required for the tender documents, generally set between 75 and 95 points, and in this embodiment, it is 85 points. The tender material price is the price at which the bidding company submits its bid, and the price deviation is the degree of deviation between the bidding company's tender price and its actual selling price.

[0078] Specifically, the process involves identifying the chapters containing the main review content of the tender documents, obtaining historical performance and financial information from the bidding qualification chapters, and determining the consistency between the historical performance and financial information to ascertain whether the bidding company has falsified its performance and financial statements. When the number of occurrences of target words in the second text information is less than the preset standard number, the complete sentences containing the target words are analyzed to determine whether the bidding company has created legal barriers to evade risk responsibility. If the analysis result is low risk, a material score representing the quality of the tender materials is determined based on the number of material characteristic words and material parameters. The degree of deviation in material price is determined based on the price of the tender materials and the bidding company's historical performance and financial information to obtain the price deviation. The review result of the tender documents is determined based on the material score and the price deviation.

[0079] Specifically, by intelligently analyzing multiple sections of the tender documents, such as qualifications, risks, materials, and pricing, a multi-dimensional and comprehensive review of the tender documents is achieved, effectively improving the comprehensiveness and depth of the review. By using natural language processing technology, key information in the tender documents, such as historical performance, financial information, risk responsibilities, and material characteristics, can be accurately identified and extracted, overcoming the subjectivity and limitations of traditional manual review. Through in-depth analysis of keywords in the risk responsibilities section, potential risk points are accurately identified, and quantitative assessments are conducted based on the degree of risk, significantly improving the accuracy and pertinence of risk identification. By comprehensively weighing material scores and price deviations, the final review result is formed, significantly improving review efficiency.

[0080] Specifically, determining whether there are false performance records based on the historical performance information and the historical financial information includes:

[0081] Calculate the correlation between the percentage fluctuation of sales performance and the percentage fluctuation of revenue data for each consecutive month in the historical performance, and compare the correlation with a preset correlation threshold to determine whether there is any false performance.

[0082] Specifically, if all correlation degrees are greater than or equal to the correlation degree threshold, it is determined that there are no false performance results; if there is a correlation degree less than the correlation degree threshold, it is determined that there are false performance results.

[0083] In the specific implementation process, the correlation threshold is 0.8. In the historical performance, the sales performance in January was 1 million and the revenue was 950,000; the sales performance in February was 1.2 million and the revenue was 1.1 million; the sales performance in March was 1.1 million and the revenue was 1.05 million; and the sales performance in April was 1.3 million and the revenue was 1.25 million. The sales performance fluctuation was calculated to be 20% in January-February, -8.3% in February-March, and 18.2% in March-April. The revenue fluctuation was 15.8% in January-February, -4.5% in February-March, and 19% in March-April. The correlation between each consecutive month, i.e., the ratio of sales performance fluctuation to revenue fluctuation, was calculated to be 0.85 in January-February, 0.82 in February-March, and 0.88 in March-April. Since all correlations are greater than or equal to the correlation threshold, it is determined that there is no false performance.

[0084] Specifically, by focusing on the dynamic fluctuation trends of performance and revenue in consecutive months, abnormal fluctuation patterns can be captured, providing a more dynamic and comprehensive assessment basis for judging false performance. By comparing the correlation between sales performance fluctuations and revenue data fluctuations, it is possible to more objectively identify whether bidders have deliberately fabricated performance information, effectively reveal potential performance fraud, and improve the accuracy and reliability of bid document review.

[0085] Please continue reading. Figure 2 ,like Figure 2As shown, it is a flowchart for determining whether a target word exists according to an embodiment of the present invention;

[0086] Specifically, determining whether the target word exists in the second text information includes:

[0087] The text of the risk and responsibility section is segmented into words to obtain several words to be determined;

[0088] By comparing the word to be determined with the preset responsibility keyword database, if the word to be determined exists in the responsibility keyword database, then it is determined that the target word exists in the second text information.

[0089] Specifically, the text in the risk and responsibility chapter is broken down into several words. The preset responsibility keyword library is a word library containing various words related to the division of risk and responsibility. When it is determined that the word to be determined is in the responsibility keyword library, it is determined that the target word exists in the second text information.

[0090] Specifically, by establishing a professional responsibility keyword database, it is possible to quickly and accurately capture potential risk signal words in the risk responsibility section, effectively overcoming the subjectivity and arbitrariness in manual review, effectively identifying hidden risk points and uncertainties in the tender documents, and using natural language processing technology to perform refined word segmentation of the text, achieving accurate deconstruction of the tender documents. It can quickly and automatically screen a large number of documents for risks, significantly improving work efficiency and reducing labor costs.

[0091] Specifically, the step of determining whether to analyze the complete sentence containing the target word based on the frequency of its occurrence compared to a preset standard frequency is as follows:

[0092] Determine the number of times the target word appears in the second text information;

[0093] Compare the number of occurrences with the preset standard number of occurrences to obtain a comparison result;

[0094] Based on the comparison results, determine whether to analyze the complete sentence containing the target word.

[0095] If the number of occurrences is less than or equal to the preset standard number of occurrences, then the complete sentence containing the target word will be analyzed.

[0096] In the specific implementation process, the preset standard number of occurrences is 1, and the number of times the target words "risk," "responsibility," and "exemption" appear in the second text information is 1. The number of occurrences is equal to the preset standard number of occurrences, and the complete sentence where the target words are located is analyzed.

[0097] Specifically, by conducting detailed statistics on the frequency of target words and comparing thresholds, risk-sensitive statements can be identified and located more accurately. The risk level can be automatically judged based on the frequency of target words. By filtering through word frequency thresholds, high-risk statements can be quickly identified, avoiding unnecessary risk misjudgments due to the appearance of individual words. This enables the quantification and hierarchical management of risks, significantly reducing the amount of calculation and processing time, and improving the overall efficiency of risk assessment.

[0098] Please continue reading. Figure 3 ,like Figure 3 As shown, it is a flowchart for determining the occurrence frequency of target words in an embodiment of the present invention;

[0099] Specifically, determining the number of times the target word appears in the second text information includes:

[0100] Determine the complete sentence in the second text information based on the target punctuation;

[0101] The number of sentences in which the target word appears in the complete sentence is determined to obtain the occurrence count.

[0102] Specifically, the target punctuation mark is the punctuation mark in the second text information that determines the complete sentence, generally a period, semicolon, etc. In this embodiment, it is a period and a semicolon. The second text information is divided into several complete sentences based on the period and semicolon in the second text information, and the number of complete sentences containing the target word in each complete sentence is determined to obtain the occurrence count.

[0103] Specifically, by accurately identifying sentence boundaries and counting the number of sentences in which target words appear in complete sentences, the analysis of text based on complete semantic units can more accurately reflect the distribution and importance of keywords in the text. Compared with the traditional word-by-word and paragraph-by-paragraph analysis method, the method of quickly locating sentence boundaries through punctuation marks significantly improves the computational efficiency of text parsing and risk identification.

[0104] Specifically, obtaining the analysis results after analyzing the complete sentence containing the target word includes:

[0105] By using natural language processing technology to perform semantic analysis on the complete sentence containing the target word, the object and manner of responsibility can be determined.

[0106] When the entity undertaking the task is the bidding company, a pre-set risk analysis model is used to conduct a risk analysis on the undertaking method to obtain the analysis results.

[0107] Specifically, natural language processing technology is used to determine the grammatical structure of complete sentences. Based on the grammatical structure, sentences related to the object and method of liability are identified. A pre-set risk analysis model is used to perform risk analysis on sentences related to the method of liability. Based on the type of liability, scope of compensation, and conditions of liability, it is determined whether the risk analysis model is met and the sentence is classified as low risk. If it is not met, the sentence is classified as high risk.

[0108] Specifically, natural language processing technology can accurately identify the subject and method of liability assumption, intelligently identify the specific object of liability assumption, and distinguish the liability boundaries of different subjects, providing a more accurate subject positioning for risk assessment. This avoids subjective judgment and fuzzy identification in traditional methods. When the object of liability assumption is the bidding company, a pre-set risk analysis model is used for targeted assessment, achieving accurate quantification and scientific evaluation of risk. This overcomes the limitations of traditional qualitative analysis. Through detailed analysis of the assumption method, risk assessment can be carried out quickly and automatically on a large number of documents, significantly improving work efficiency.

[0109] Specifically, the calculation of material scores based on the number of material feature terms, material parameters, and the corresponding number of standard feature terms and standard parameters includes:

[0110] The feature score is determined based on the number of material feature words and the number of standard feature words;

[0111] The material score is determined based on the material parameters and the standard parameters.

[0112] The material score is calculated by adding the feature score and the material score together with different weights.

[0113] In the specific implementation process, the standard feature words are 3, the standard parameter is 80HB, and the actual number of material feature words in the bidding materials is 2, that is, the feature score is 2 / 3*100=66.67 points. The actual material parameter in the bidding materials is 80HB, that is, 80 / 80*100=100 points. The feature score weight is 0.1, and the material parameter score weight is 0.9. Therefore, the calculated material score is 66.67*0.1+87.5*0.9=26.67+52.5=85.417 points.

[0114] Specifically, by simultaneously considering the number of material characteristic terms and material parameters, it can more comprehensively reflect the overall quality of materials. It adopts two dimensions, feature scoring and material scoring, and assigns different weights. The scoring weights can be flexibly adjusted according to the characteristics of different industries and projects, which has good adaptability and customization capabilities. It also achieves accurate quantification of material quality, overcomes the subjectivity and arbitrariness of traditional qualitative assessment, and improves the accuracy and objectivity of the assessment.

[0115] Specifically, determining the price deviation based on the historical performance information, the historical financial information, and the price of the tendered materials includes:

[0116] Calculate the ratio of total sales volume in the historical performance information to total revenue in the historical financial information to obtain the sales price;

[0117] Calculate the ratio of the price of the tendered materials to the selling price to obtain the price deviation.

[0118] In the specific implementation process, the historical total sales volume in the tender document is 10,000 units, the historical total revenue is 50 million yuan, the historical sales price is 50 million yuan / unit (10,000 units), and the price of the tendered materials is 480 yuan / unit. Therefore, the price deviation is 480 divided by 500, which is 0.96.

[0119] Specifically, by comprehensively considering historical performance, financial information, and current bid prices, the system significantly improves the comprehensiveness and accuracy of price review. The method of calculating the ratio of bid material prices to historical sales prices can sensitively capture price deviations, providing a scientific quantitative indicator for identifying unreasonable quotations, effectively preventing potential commercial fraud and unreasonable pricing risks, and providing effective decision support for bid evaluation.

[0120] Specifically, determining the review result based on the deviation between the material score and the price includes:

[0121] The ratio of the material score to the price deviation is calculated to obtain the review result.

[0122] In the specific implementation process, the material score is 85.417 points and the price deviation is 0.96. Then, the ratio of the material score of 85.417 to the price deviation of 0.96 is calculated as 85.417 ÷ 0.96, which is 88.98.

[0123] Specifically, by dividing the material score by the price deviation, the system can automatically review bid documents based on material quality and price risks, achieving intelligent calculation of bid review results and significantly improving review efficiency and accuracy.

[0124] Please continue reading. Figure 4 ,like Figure 4 The diagram shown is a system framework diagram of the automatic bid document review system based on natural language according to an embodiment of the present invention.

[0125] This invention also provides a natural language-based automatic bid document review system applicable to any of the above-mentioned automatic review methods, comprising:

[0126] The chapter determination module is used to identify the target chapters in the tender documents to be processed.

[0127] The text analysis module, which is connected to the chapter determination module, is used to analyze the first text information in the bidding qualification chapter to obtain historical performance information and historical financial information.

[0128] A fraud detection module, which is connected to the text analysis module, is used to determine whether there is fraudulent performance based on the historical performance information and the historical financial information.

[0129] The risk assessment module, which is connected to the chapter determination module and the fraud determination module, is used to obtain the second text information in the risk responsibility chapter when it is determined that there is no fraudulent performance, and to determine whether to analyze the complete sentence where the target word is located based on the number of times the target word appears and the preset standard number of times when the target word is present in the second text information, and to obtain the analysis results.

[0130] The material scoring module, which is connected to the chapter determination module and the risk assessment module, is used to determine the number of material feature words and material parameters in the third text information and calculate the material score when the analysis result is low risk.

[0131] The deviation calculation module, which is connected to the chapter determination module and the text analysis module, is used to compare the material score with the material score threshold, and determine the price deviation when the material score is greater than the material score threshold.

[0132] The review module, which is connected to the material scoring module and the deviation calculation module, is used to determine the review result based on the material score and the price deviation.

[0133] Specifically, the natural language-based automatic bid review system provided in this embodiment can run the above-mentioned natural language-based automatic bid review method and achieve the same technical effect, which will not be repeated here.

[0134] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0135] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for automatic evaluation of tender documents based on natural language, characterized in that, include: Identify the target sections in the tender documents to be processed, including the material information section, the tender qualification section, the risk and responsibility section, and the tender price section; Analyze the first text information in the bidding qualification section to obtain historical performance information and historical financial information; Determine whether there are any false performance records based on the historical performance information and the historical financial information; When it is determined that there is no false performance, the second text information in the risk responsibility section is obtained, and it is determined whether there is a target word in the second text information. When it is determined that there is a target word, it is determined whether to analyze the complete sentence where the target word is located based on the number of times the target word appears and the preset standard number of times. The analysis results are obtained after analyzing the complete sentence containing the target word; When the analysis result is low risk, obtain the third text information in the material information section; Determine the number of material feature words and material parameters in the third text information, and calculate the material score based on the number of material feature words, material parameters, and the corresponding number of standard feature words and standard parameters; Compare the material score with the material score threshold. If the material score is greater than the material score threshold, obtain the bid material price in the bid quotation section. Determine the price deviation based on the historical performance information, the historical financial information and the bid material price. The review result is determined based on the material score and the price deviation. The determination of whether there is false performance based on the historical performance information and the historical financial information includes: Calculate the correlation between the sales performance fluctuation ratio and the revenue data fluctuation ratio in each consecutive month of the historical performance, and compare the correlation with a preset correlation threshold to determine whether there is false performance. The calculation of material scores based on the number of material feature terms, material parameters, and the corresponding number of standard feature terms and standard parameters includes: The feature score is determined based on the number of material feature words and the number of standard feature words; The material score is determined based on the material parameters and the standard parameters. The material score is calculated by adding the feature score and the material score with different weights. The determination of price deviation based on the historical performance information, the historical financial information, and the price of the tendered materials includes: Calculate the ratio of total sales volume in the historical performance information to total revenue in the historical financial information to obtain the sales price; Calculate the ratio of the price of the tendered materials to the selling price to obtain the price deviation. The determination of the review result based on the material score and the price deviation includes: The ratio of the material score to the price deviation is calculated to obtain the review result.

2. The natural language based bid document automatic review method of claim 1, wherein, Determining whether the target word exists in the second text information includes: The text of the risk and responsibility section is segmented into words to obtain several words to be determined; By comparing the word to be determined with the preset responsibility keyword database, if the word to be determined exists in the responsibility keyword database, then it is determined that the target word exists in the second text information.

3. The method for automatic evaluation of tender documents based on natural language according to claim 2, characterized in that, The process of determining whether to analyze the complete sentence containing the target word based on the frequency of its occurrence compared to a preset standard frequency is as follows: Determine the number of times the target word appears in the second text information; Compare the number of occurrences with the preset standard number of occurrences to obtain a comparison result; Based on the comparison results, determine whether to analyze the complete sentence containing the target word. If the number of occurrences is less than or equal to the preset standard number of occurrences, then the complete sentence containing the target word will be analyzed.

4. The method for automatic evaluation of tender documents based on natural language according to claim 3, characterized in that, Determining the frequency of occurrence of the target word in the second text information includes: Determine the complete sentence in the second text information based on the target punctuation; The number of sentences in which the target word appears in the complete sentence is determined to obtain the occurrence count.

5. The method for automatic evaluation of tender documents based on natural language according to claim 4, characterized in that, The process of obtaining analysis results after analyzing the complete sentence containing the target word includes: By using natural language processing technology to perform semantic analysis on the complete sentence containing the target word, the object and manner of responsibility can be determined. When the entity undertaking the task is the bidding company, a pre-set risk analysis model is used to conduct a risk analysis on the undertaking method to obtain the analysis results.

6. A system using the natural language-based automatic bid evaluation method according to any one of claims 1-5, characterized in that, include: The chapter determination module is used to identify the target chapters in the tender documents to be processed. The text analysis module, which is connected to the chapter determination module, is used to analyze the first text information in the bidding qualification chapter to obtain historical performance information and historical financial information. A fraud detection module, which is connected to the text analysis module, is used to determine whether there is fraudulent performance based on the historical performance information and the historical financial information. The risk assessment module, which is connected to the chapter determination module and the fraud determination module, is used to obtain the second text information in the risk responsibility chapter when it is determined that there is no fraudulent performance, and to determine whether to analyze the complete sentence where the target word is located based on the number of times the target word appears and the preset standard number of times when the target word is present in the second text information, and to obtain the analysis results. The material scoring module, which is connected to the chapter determination module and the risk assessment module, is used to determine the number of material feature words and material parameters in the third text information and calculate the material score when the analysis result is low risk. The deviation calculation module, which is connected to the chapter determination module and the text analysis module, is used to compare the material score with the material score threshold, and determine the price deviation when the material score is greater than the material score threshold. The review module, which is connected to the material scoring module and the deviation calculation module, is used to determine the review result based on the material score and the price deviation.

Images